Validation Study on an Information Driven Library Design Strategy

Abstract

A new method is introduced for performing reagent selection for chemical library design based on topological (2D) pharmacophore fingerprints. Optimal reagent selection is achieved by optimising the Shannon entropy of the 2D pharmacophore distribution for the reagent set. The method, termed ProSAR, is therefore expected to enumerate compounds that could serve as a good starting point for deriving a structure activity relationship (SAR) in combinatorial library design. The main goal for current study is to validate this methodology by applying it on several library design examples where the active compounds were already known and comparing the performance of ProSAR libraries with random libraries and traditional diversity based libraries. The results show that ProSAR libraries generally have better pharmacophore coverage than libraries coming from other design strategies. The effectiveness of generating active compounds for the designed library is also evaluated by first doing a similarity search against GVKBio database with library compounds as query structures, then comparing the number of retrieved active compounds for different libraries. The results demonstrate that in most of cases, ProSAR libraries retrieve more active compounds than other libraries. The ProSAR strategy is further expanded to include product property profiles for aqueous solubility, hERG risk assessment etc. in the optimisation process so that the reagent pharmacophore diversity and the product property profile are optimised simultaneously via a genetic algorithm. The validation study results show that by using the ProSAR methodology, the designed libraries can achieve good pharmacophore coverage and product property profile simultaneously.